<![CDATA[Voltage sags are frequent disturbances in industrial power systems that can disrupt system operations and cause equipment malfunctions. The proposed framework integrates Random Matrix Theory (RMT) to identify disturbance patterns. It evaluates the severity, vulnerabilities, and operational impact of voltage sag events using the Information Technology Industry Council (ITIC) curve. This research uses event data, disturbance type, associated equipment, disturbance duration, and three-phase voltage measurements to predict the temporal evolution of ITIC conditions in pattern disturbance dynamics. Deep learning models, namely Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU), are then employed to predict the temporal evolution of ITIC conditions. Based on power metering unit measurements, the observed voltage variations were non-linear, yet the RMT stability index (Ψ) remained within ITIC tolerance limits. The severity of stability disturbances was successfully evaluated, and the GRU model demonstrated superior predictive performance compared to the LSTM model. Consequently, the industry requires an AVC system—aligned with the combined stability-severity-risk paradigm and the prediction results—to effectively mitigate voltage compensation risks through precise AVC operation. These findings demonstrate that integrating RMT-based fault analysis, ITIC-based severity assessment, and deep learning-based prediction offers a more systematic and predictive approach to voltage sag assessment than relying solely on empirical evaluation. Consequently, this enables more accurate determination of AVC installation requirements, thereby effectively mitigating faults. The implication is a shift in how AVC requirements are assessed—moving from a reactive to a predictive approach—thereby reducing the risk of inadequate or unnecessary compensation.]]>
Copyrights © 2026
